AI embedded in the EHR helps prevent adverse medication interactions

Photo: Morsa Images/Getty Images

Artificial intelligence has stepped in to help complete the medication information in the medical record for patients who can't remember what drugs they are taking, let alone the dose. 

This is most of us, according to Rebecca Sulfridge, a clinical pharmacist and emergency specialist at Covenant HealthCare in Michigan. Most patients have gaps in their medication history. 

Technicians hear, "'I take three pills. One is a little white one. One is a pink one,'" or "'I know I take two blood pressure meds and one to sleep at night,'" in face-to-face talks with patients, according to Sulfridge. Even those patients who know the convoluted names of all the drugs they're taking may not know the dose.

It's Sulfridge's job to reconcile the medication medical record in the emergency room.

"The heart and soul lies with the medical history," Sulfridge said.

Improvements have been made through artificial intelligence embedded in the health system's electronic health record.

During HIMSS21, Sulfridge will talk about how to Improve Emergency Department Patient, Staff Safety with an AI-Enabled Medication History, on Tuesday, August 10, from 1-2 p.m. in the Venetian, Murano 3204.

Errors in medication information is a safety concern, especially in the emergency room when patients arrive for a health issue and clinicians need to know what other prescriptions are being taken that could create an adverse interaction with a new drug.

Generally, gaps in information mean calling the pharmacy, or the prescribing physician, or the patient's caregiver, or all three.

In one case, a patient who had recently undergone a kidney transplant came in with an infection. Their medication list was at least six months old, Sulfridge said, but the clinicians knew this patient had a kidney transplant after that and that new medications were likely prescribed, but not on the list.

"They ended up calling the daughter or the caregiver," Sulfridge said. "It's important to prevent adverse effects from medication errors."

In 2014, the AI pilot program was present to health system executives. It took convincing because of uncertainty over the cost savings of such a program.

Hospitals can't bill insurers for taking a medication history, Sulfridge said. Plus, it created new positions.

"It's difficult to sell a program to executives when it appears it costs them money and manpower," Sulfridge said.

Executives were given a cost-benefit analysis of error avoidance with an estimated savings of $6 million a year.

Software vendor DrFirst embedded software in the Epic EHR program to upload claims data on medication. Clinicians were able to see which prescriptions had not only been prescribed, but which ones were being filled.

"With our software, I can click on Lipitor and see that they take 40 mms at bedtime," Sulfridge said.

During the pandemic, in which the health system experienced three surges, the program was invaluable, Sulfridge said. This is because technicians, whose job it is to take the medication histories, weren't allowed in the rooms, since there wasn't enough personal protection equipment to go around. Nurses did their best to collect the data, but they had much else to do as well.

"Without that, I think we would not have been able to do medication histories," Sulfridge said.

COVID-19 presented unique challenges. Early on during the pandemic, the ER was eerily quiet, Sulfridge said. People were afraid to come into the hospital. 

"We were seeing people coming in two days after they had a heart attack," Sulfridge said.

She doesn't know why, but during COVID-19, productivity in medication history increased 14-15%.

The future includes software on allergy information and price transparency to see what medication is covered, and what's not covered, for a particular patient, she said.

"It's been really, really interesting to see how changes to workflow came out of COVID, and see the changes, the increase in productivity and accuracy and with AI," she said.

Twitter: @SusanJMorse
Email the writer: susan.morse@himssmedia.com